Prediction time

[vladimircape]

Hi, very interesting approach.
Can you provide some estimates of time performance for predict image classification.
For example how much time takes to predict 28x28 RGB image, it will be great to see comparison by VGG16
Thanks

[kk7nc]

Thank you for your question,
The time for training and prediction depends on many factors such as the size of the dataset, how many models you want to generate (CNN, RNN, and DNN), your hardware, and number of epochs,
I think if you create 3 DNNs and 3 CNNs with mid-level hardware (one GPU), it will be learned models very fast (2-3 hours).
Best regards,
Kamran

[vladimircape]

Thanks for answer.
Time for training is not important to me, most interesting time for a prediction. I think it will not depend on size of dataset and number of epochs.
Right now i use VGG19 to make a prediction on Video(60fps) and on my machine with Geforce 1060, it takes 20ms to make a prediction, for InceptionV1 as I remember 150ms.
It's very interesting to me to make approximate comparison of time for prediction, to know if you RMDL is suitable by performance, for my task.
Thanks. And great work

[kk7nc]

Prediction time is very fast and even if you use several models it could be real-time,
best regards,
Kamran

[vladimircape]

ok, i will try and provide result later